Abstract
Background:
Cytotoxic agents have a narrow therapeutic window. A high percentage of some of them is renally excreted in unchanged form. Accurate assessment of an individual's glomerular filtration rate (GFR) helps to predict the pharmacokinetic behavior of those drugs more precisely. GFR calculations, however, have their limitations.
Objective:
To establish a more accurate calculation of renal function over a broad range of constitutive GFR values.
Methods:
Patients with cancer were included in the analysis. Serum levels of cystatin C, creatinine, urea, albumin, and β-trace protein were measured, and GFR was calculated by 8 mathematical formulas. The results were compared with creatinine clearance (CrCI) calculated from timed urine specimens.
Results:
One hundred two patients were evaluated: median age, 57.5 years (range 20–91); females, 52; males, 50; and mean urinary CrCI, 77.0 mL/min. The bias (mean percentage error) was -2% and the precision (mean absolute percentage error) was 23% for the Modification of Diet in Renal Disease (MDRD) estimation of GFR. All equations significantly overestimated CrCI in patients with measured clearance less than 50 mL/min (p <0.05), with the exception of the modified Salazar-Corcoran formula. All equations underestimated CrCI in patients with measured clearance greater than 100 mL/min. The Wright formula was the least biased and most precise (–5%, 16%, respectively). In patients with measured CrCI 50–100 mL/min, the MDRD calculation had a bias of -4% and a precision of 17%. The Jelliffe and Larsson equations were associated with significant sex bias (p <0.05).
Conclusions:
These observations suggest that individual GFR values over a broad range cannot be calculated accurately enough with only one selected formula. It may be useful to classify renal function of patients with cancer according to the novel algorithm by using MDRD first and then to subsequently calculate GFR in higher and lower ranges with the Wright and modified Salazar-Corcoran formulas, respectively. This algorithm should be validated using larger numbers of patients.
Get full access to this article
View all access options for this article.